Update README.md

Cosmos3-Edge-int4-convrot.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:416c256176f52f33768d5a551a090b63598b0ce9a26f872174543df5bf8631de
-size 3129557536

Cosmos3-Nano-int4-convrot.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:2e94116fc798813b64c6922d746f5214b00eb1d0e12ff3ffb8433a329450ea50
-size 12382144496

Cosmos3-Super-Image2Video-4Step-int4-convrot.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:af234e8008f270fd1adf311d1d1a2ecaba396f728735097f633796d8a955751b
-size 46718563752

Cosmos3-Super-Image2Video-int4-convrot.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:969d419cf0dd1f152f1188eeb46eb3994639ee956d8691d60c2606e4e12c9985
-size 46718563752

Cosmos3-Super-int4-convrot.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:ed496d2cd525bb12409bb8bc83f66e83c405a9fc930ca70bc2d4ea745e93083c
-size 46762181072

README.md

@@ -1,133 +1,64 @@
 ---
-license: other
+license: openmdw-1.1
 license_name: openmdw-1.1
-license_link: https://openmdw.ai/license/1-1/
 tags:
-  - text-to-video
-  - image-to-video
-  - quantized
-  - int8
-  - int4
-  - comfyui
-  - cosmos3
+- text-to-video
+- image-to-video
+- quantized
+- int8
+- comfyui
+- cosmos3
 ---
 
-# Cosmos3 quantized transformers
+# Cosmos3 ConvRot (INT8)
 
-Weight-only quantized transformers of the NVIDIA Cosmos3 models, for use with
-[ComfyUI-Cosmos3](https://github.com/RyukoMatoiFan/ComfyUI-Cosmos3). Each file is one quantized
-transformer; take the VAE, tokenizer(s) and `config.json` from the matching official
-`nvidia/Cosmos3-*` repo.
+INT8-ConvRot quantized transformers of the NVIDIA Cosmos3 models, for use with ComfyUI-Cosmos3
+(https://github.com/RyukoMatoiFan/ComfyUI-Cosmos3). Each file is one quantized transformer; take the
+VAE, tokenizer(s) and config.json from the matching official nvidia/Cosmos3-* repo.
 
-Two formats:
-
-- **int8** — weight-only INT8, per-output-channel scale + a group-wise Hadamard rotation (ConvRot).
-- **int4** — MLP weights INT4 (GPTQ-calibrated with a ConvRot Hadamard rotation, packed as AWQ
-  W4A16); attention kept INT8; the excluded layers (below) stay bf16. Larger quantization error than
-  int8 (at a fixed seed the output diverges from bf16 more than int8 does); about a third of the bf16
-  size.
-
-Both are **weight-only**: activations stay bf16, so this lowers memory/download size, not compute
-speed. int4 runs at about bf16 speed — the per-forward dequantize and un-rotate add a little.
-
-## Files and sizes
-
-| Model | bf16 | int8 | int4 |
-|-------|------|------|------|
-| `Cosmos3-Nano` | 30 GB | **16.5 GB** | **12.4 GB** |
-| `Cosmos3-Super` | 128 GB | **65.7 GB** | **46.8 GB** |
-| `Cosmos3-Super-Image2Video` | 128 GB | **65.6 GB** | **46.7 GB** |
-| `Cosmos3-Super-Image2Video-4Step` | 128 GB | **65.6 GB** | **46.7 GB** |
-| `Cosmos3-Edge` | 6.7 GB | **3.9 GB** | **3.0 GB** |
-
-File names: `Cosmos3-<name>-int8-convrot.safetensors` and `Cosmos3-<name>-int4-convrot.safetensors`.
-int4 and int8 are provided for every model.
-
-**Prompting note (Edge):** `Cosmos3-Edge` is trained on JSON-structured prompts and is less robust to
-plain text than the larger Nano/Super. Plain text usually works, but on some detailed scenes (notably
-reflective surfaces) it can produce flare/pulsation artifacts; wrapping the text as
-`{"temporal_caption": "<your prompt>"}` avoids them. This is a base-model property, not a quantization
-effect (it shows in bf16 too).
-
-## Runtime footprint
-
-With ComfyUI's dynamic VRAM the transformer is streamed from host RAM, so the GPU holds only the
-activations. Measured at 832×480, 93 frames, at the minimum VRAM budget (maximum streaming):
-
-| Model | Min VRAM | RAM (bf16 / int8 / int4) |
-|-------|----------|---------------------------|
-| `Cosmos3-Edge` | ≈6 GB | 14 / 7 / 7 GB |
-| `Cosmos3-Nano` | ≈7 GB | 58 / 21 / 20 GB |
-| `Cosmos3-Super` (t2v & i2v) | ≈8–9 GB | 240 / 67 / 63 GB |
-
-Min VRAM is the activation floor (set by resolution × frame count, not the weight format). RAM is the
-peak host memory — larger than the file on disk (staging + overhead), and bf16 peaks near twice the
-weight size. RAM and VRAM trade off: giving the GPU more VRAM holds more weights on-card and lowers the
-RAM figure. The Super-family int4 checkpoints fit a 64 GB host (≈63 GB); int8 needs a little more
-(≈67 GB).
+| File | Source repo | Size (bf16 → int8) | Modalities |
+|------|-------------|--------------------|------------|
+| `Cosmos3-Nano-int8-convrot.safetensors` | nvidia/Cosmos3-Nano | 29 → 16 GB | t2v / i2v / audio, fps 24 |
+| `Cosmos3-Super-int8-convrot.safetensors` | nvidia/Cosmos3-Super | 120 → 62 GB | t2v / i2v / audio, fps 24 |
+| `Cosmos3-Super-Image2Video-int8-convrot.safetensors` | nvidia/Cosmos3-Super-Image2Video | 120 → 62 GB | i2v, fps 16 |
+| `Cosmos3-Super-Image2Video-4Step-int8-convrot.safetensors` | nvidia/Cosmos3-Super-Image2Video-4Step | 120 → 62 GB | i2v, DMD2 4-step, cfg 1 |
+| `Cosmos3-Edge-int8-convrot.safetensors` | nvidia/Cosmos3-Edge | 6.3 → 3.7 GB | t2v / i2v, nemotron_dense |
 
 ## Usage
 
-1. Download the official `nvidia/Cosmos3-<name>` into `ComfyUI/models/cosmos3/<name>/`.
-2. In its `transformer/` folder, delete the bf16 shards and `*.index.json`, then put the quantized
-   file there renamed to `diffusion_pytorch_model.safetensors`. Keep the official `config.json`,
-   `vae/`, `text_tokenizer/`, `sound_tokenizer/`.
-3. Load with the Cosmos3 Loader (`weight_dtype = default`); the loader reads the format from the
-   checkpoint metadata. Requires comfy-kitchen (int8 from ComfyUI >= 0.27) and the latest
-   ComfyUI-Cosmos3 (int4 needs the ConvRot-aware loader).
+1. Download the official nvidia/Cosmos3-<name> into ComfyUI/models/cosmos3/<name>/.
+2. In its transformer/ folder, delete the bf16 shards and *.index.json, then put the int8 file there
+   renamed to diffusion_pytorch_model.safetensors. Keep the official config.json, vae/,
+   text_tokenizer/, sound_tokenizer/.
+3. Load with the Cosmos3 Loader (weight_dtype = default). Requires a ComfyUI with native INT8
+   (>= 0.27, comfy-kitchen + Triton) and the latest ComfyUI-Cosmos3.
 
-## How these were quantized
+## Quantization
 
-Reproducible in method, not bit-for-bit (the calibration set and RNG vary per run). Both formats keep
-the same **escape set in bf16**: `proj_in`, `proj_out`, `time_embedder`, `audio_proj`,
-`modality_embed`, the embeddings, and all norms and biases.
+Format: comfy_quant int8_tensorwise with convrot=true (weight-only INT8; quantize_input=false, so
+activations stay in the bf16 compute dtype). Weights are per-output-channel (row-wise) symmetric INT8
+with a float32 weight_scale of shape [out_features, 1]; a group-wise Hadamard rotation (ConvRot, group
+size 256) is applied to each weight before quantization. Excluded layers stay bf16 (see below).
 
-### int8 — every linear
-
-Weight-only symmetric INT8, per-output-channel scale (`weight_scale`, float32, `[out, 1]`), with a
-group-wise Hadamard rotation (ConvRot, **group 256**) applied before quantization and undone at load.
-comfy_quant tag `int8_tensorwise`, `convrot=true`. No calibration: at 8-bit the per-channel scale and
-the rotation keep the round-to-nearest error small — error feedback (GPTQ) is only needed at int4.
-Produced with
-[convert_to_quant](https://github.com/silveroxides/convert_to_quant):
+Produced with convert_to_quant (ctq): https://github.com/silveroxides/convert_to_quant
 
+    # 1. consolidate the sharded diffusers transformer into one safetensors, then:
     ctq -i transformer_bf16.safetensors -o out_int8_convrot.safetensors \
         --int8 --scaling_mode row --simple --convrot --convrot-group-size 256 \
         --comfy_quant --save-quant-metadata --cosmos3 --device cuda --low-memory
 
-### int4 — MLP → int4, attention → int8
-
-Round-to-nearest INT4 — even with the ConvRot rotation — leaves visible artifacts on these models, so
-the MLP path uses **GPTQ error compensation on real activations**. Steps:
-
-1. **Capture activations.** Run genuine denoising at 832×480, 93 frames with the model's normal
-   sampler (t2v / base-i2v: 35 steps, cfg 6, `uni_pc_bh2`; 4-step model: 4 steps, cfg 1, `euler`) over
-   ≈4 prompts, with a forward pre-hook on every target linear. Keep a reservoir of up to **4096** rows
-   per layer (random replacement beyond that). The understanding tower sees the text prefill; the
-   generation tower sees every denoising step.
-2. **Rotate (ConvRot).** Apply a Sylvester block-Hadamard — symmetric, orthonormal, **group 32** — to
-   both the weight and the captured activations of each MLP linear.
-3. **GPTQ.** Quantize the rotated weight to symmetric INT4 (codes −8…7, per-(row, group-32) scale).
-   Hessian `H = XᵀX · 2/N` from the rotated activations; diagonal damping raised through
-   `{0.01, 0.03, 0.1, 0.3, 1, 3} × mean(diag)` until the Cholesky factors; columns processed in blocks
-   of **128** with per-column error feedback into the not-yet-quantized columns; plain round-to-nearest
-   only if damping never succeeds.
-4. **Attention → int8** (row-wise, `[out, 1]` scale). INT4 on attention produces visible artifacts.
-5. **Pack** the INT4 codes into comfy-kitchen's **AWQ W4A16** layout; the loader un-rotates each group
-   at dequant.
+--simple skips learned rounding (ConvRot handles the weight outliers; learned rounding adds only
+~0.001 latent cosine for much more compute).
 
-Layer counts follow the checkpoint's `config.json` — e.g. `Cosmos3-Super-Image2Video` packs 384 MLP
-linears (INT4) + 512 attention linears (INT8).
+--cosmos3 is a layer-exclusion preset kept in bf16, in addition to ctq's base avoid-list
+(norm, bias, embed_tokens, lm_head, k_norm, q_norm): proj_in, proj_out, time_embedder, audio_proj,
+action, modality_embed. To reproduce, add a cosmos3 entry to convert_to_quant/constants.py
+MODEL_FILTERS with exclude = AVOID_KEY_NAMES and highprec = that list.
 
-**Execution.** Both formats run as an explicit dequantize-then-bf16 matmul (unpack the 4-bit weights
-to bf16, un-rotate, multiply in bf16). This is weight-only regardless: there is no int4-weight ×
-bf16-activation tensor-core op on any GPU (Hopper or Blackwell — Blackwell's FP4 cores are for NVFP4,
-both operands 4-bit, not this W4A16 layout), so a W4A16 kernel would dequantize internally too, with no
-compute speedup either way. We dequantize explicitly instead of calling comfy-kitchen's
-`gemv_awq_w4a16`, which is non-deterministic (atomic accumulation jitters the video frame-to-frame) and
-numerically off on these shapes.
+Checked per model against bf16: Nano latent cosine 0.986; others visually indistinguishable.
+4-bit (NVFP4) is not provided: it requires a Blackwell GPU (SM >= 10.0) for both conversion and
+inference.
 
 ## License
 
-Derived from NVIDIA Cosmos3 checkpoints; the [OpenMDW-1.1](https://openmdw.ai/license/1-1/) license
-applies (same as the upstream `nvidia/Cosmos3-*` models).
+Derived from NVIDIA Cosmos3 checkpoints; the OpenMDW-1.1 License applies.